Table 3 Comparison of segmentation AP.
From: Deep learning application of vertebral compression fracture detection using mask R-CNN
Models/fractures | L1 | L2 | L3 | L4 | T11 | T12 | Overall mAP |
|---|---|---|---|---|---|---|---|
Mask R-CNN | 0.80 | 0.79 | 0.66 | 0.32 | 0.10 | 0.78 | 0.58 |
Cascade mask R-CNN | 0.63 | 0.58 | 0.33 | 0.24 | 0.07 | 0.52 | 0.40 |
YOLOACT | 0.70 | 0.66 | 0.67 | 0.30 | 0.22 | 0.67 | 0.54 |
YOLOv5-s | 0.72 | 0.78 | 0.72 | 0.24 | 0.09 | 0.68 | 0.54 |
YOLOv5-m | 0.73 | 0.77 | 0.67 | 0.25 | 0.02 | 0.68 | 0.52 |
YOLOv5-l | 0.78 | 0.82 | 0.69 | 0.13 | 0.07 | 0.75 | 0.54 |
YOLOv5-x | 0.78 | 0.83 | 0.78 | 0.54 | 0.12 | 0.72 | 0.58 |
